Ground Fault Protection Trigerring in Server Room

No, it triggers when the difference in current flow in the hot and neutral exceeds a permitted value. This could be because the current is leaking to the ground, or anywhere else.

It might be that the leakage in the PSU's is exceeding the trip current.

In UK (and I think all the EU), RCDs (GFIs) to protect against electrocution are 30mA trip (which means they must trip at somewhere between 15mA and 30mA). General recommendation is that max design leakage is 25% of the trip rating, i.e. 7.5mA. Each piece of Class I (earthed) IT equipment is allowed to leak up to 0.75mA through its RFI components. That gives a maximum of

10 pieces of Class I (earthed) IT equipment per 30mA RCD. In terms of desktop PCs, that's about 5 (each one consisting of a base unit and monitor, which are Class I). (However, we would not normally use RCD protection in server rooms at all.)

Now, all these figures are going to be different in the US and you need to perform a similar calculation. In particular, your GFI's are normally 5mA, which I would imagine is going to severely limit the number of pieces of Class I (earthed) IT equipment per GFI.

What he said. Some (most now) power supplies are power factor corrected. They have filters that cause a small amount of leakage current to flow on the ground. Sum enough of these together and you trip a GFI (or even an RCD for really large gear). It appears your two machines each have power supplies that are close to tripping the GFI with one power supply connected, and will trip when both are connected.

Charles Perry P.E.

On 06 Jul 2008 07:10:25 GMT Andrew Gabriel wrote: | In article , | "Will" writes: |> Our computer server room outlets for our power distribution units (PDUs) are |> protected by ground fault protection. As I understand it, a GFP will |> trigger when it senses current on the ground, and when it does then the GFP | | No, it triggers when the difference in current flow in the hot | and neutral exceeds a permitted value. This could be because | the current is leaking to the ground, or anywhere else. | |> shuts off the circuit. We tried today to rewire some things in our |> server room, and the GFP kept triggering, and I want help understanding |> possible causes. |> |> Each server has two power supplies. In an effort to minimize cabling, we |> purchased from Dell some cables that plug into the PDU using Nema 5-15P, and |> split to two different IEC C14 plugs that go to the server power supplies. |> It appears that use of this power cord to two power supplies is what is |> causing the ground faults. We can cure the GFP fault simply by leaving |> just one of the two plugs in each server. Does someone have an explanation |> as to why the GFP would fault when both power supplies are served from the |> same plug? | | It might be that the leakage in the PSU's is exceeding the trip | current. | | In UK (and I think all the EU), RCDs (GFIs) to protect against | electrocution are 30mA trip (which means they must trip at | somewhere between 15mA and 30mA). General recommendation is that | max design leakage is 25% of the trip rating, i.e. 7.5mA. Each | piece of Class I (earthed) IT equipment is allowed to leak up | to 0.75mA through its RFI components. That gives a maximum of | 10 pieces of Class I (earthed) IT equipment per 30mA RCD. In | terms of desktop PCs, that's about 5 (each one consisting of | a base unit and monitor, which are Class I). (However, we would | not normally use RCD protection in server rooms at all.) | | Now, all these figures are going to be different in the US | and you need to perform a similar calculation. In particular, | your GFI's are normally 5mA, which I would imagine is going to | severely limit the number of pieces of Class I (earthed) IT | equipment per GFI.

There are 2 levels of GFI protection used in the US, 6ma and 30ma. For the

6ma level, the allowed range is 2ma (most not trip below that) and 6ma (must trip any higher than this) with the nominal supposedly at 5ma (where most will trip). We do have the 30ma level for industrial (not home) use, which might be suitable for a server room where GFI is somehow required. I do not know the range for that.

On Sat, 5 Jul 2008 19:01:09 -0700 Will wrote: | Our computer server room outlets for our power distribution units (PDUs) are | protected by ground fault protection. As I understand it, a GFP will | trigger when it senses current on the ground, and when it does then the GFP | shuts off the circuit. We tried today to rewire some things in our | server room, and the GFP kept triggering, and I want help understanding | possible causes.

Do your local requirements actually specify that you must have GFI protection? If so, do you know if the 6ma level or 30ma level is specified? What is the level installed?

If you don't need to have this kind of protection, then maybe it's better to just avoid it.

| Each server has two power supplies. In an effort to minimize cabling, we | purchased from Dell some cables that plug into the PDU using Nema 5-15P, and | split to two different IEC C14 plugs that go to the server power supplies. | It appears that use of this power cord to two power supplies is what is | causing the ground faults. We can cure the GFP fault simply by leaving | just one of the two plugs in each server. Does someone have an explanation | as to why the GFP would fault when both power supplies are served from the | same plug?

You are using "double cords" plugging two computers into one outlet. Does each outlet have its own separate GFI protection (such as being fed by a separate GFI type breaker)? If so, the other explanations about marginal leakage would cover it. And those filters could be leaking extra to ground if your power quality is bad, and there's lots of noise on the line.

FYI, GFI can be tripped with high frequency energy. I've done so with 146 MHz that was apparently resonating in common mode on the power wires the GFI was on. Common mode is what GFI would trip on.

| Perhaps the power supplies start up at different times, and there is some | kind of reverse current associated with the startup?

That by itself would not trigger the GFI protection.

I wondered about this, is there actually a leakage current or is th current phase shifted to the extent the GFCI "thinks" there is an imbalance?

: >>> shuts off the circuit. We tried today to rewire some things in our : >>> server room, and the GFP kept triggering, and I want help understanding : >>> possible causes. : >>>

: >>> Each server has two power supplies. In an effort to minimize cabling, : >>> we : >>> purchased from Dell some cables that plug into the PDU using Nema

5-15P, : >>> and : >>> split to two different IEC C14 plugs that go to the server power : >>> supplies. : >>> It appears that use of this power cord to two power supplies is what is : >>> causing the ground faults. We can cure the GFP fault simply by leaving : >>> just one of the two plugs in each server. Does someone have an : >>> explanation : >>> as to why the GFP would fault when both power supplies are served from : >>> the : >>> same plug? : >> It might be that the leakage in the PSU's is exceeding the trip : >> current. : >>

: >> In UK (and I think all the EU), RCDs (GFIs) to protect against : >> electrocution are 30mA trip (which means they must trip at : >> somewhere between 15mA and 30mA). General recommendation is that : >> max design leakage is 25% of the trip rating, i.e. 7.5mA. Each : >> piece of Class I (earthed) IT equipment is allowed to leak up : >> to 0.75mA through its RFI components. That gives a maximum of : >> 10 pieces of Class I (earthed) IT equipment per 30mA RCD. In : >> terms of desktop PCs, that's about 5 (each one consisting of : >> a base unit and monitor, which are Class I). (However, we would : >> not normally use RCD protection in server rooms at all.) : >>

: >> Now, all these figures are going to be different in the US : >> and you need to perform a similar calculation. In particular, : >> your GFI's are normally 5mA, which I would imagine is going to : >> severely limit the number of pieces of Class I (earthed) IT : >> equipment per GFI. : >>

: >> -- : >> Andrew Gabriel : >> [email address is not usable -- followup in the newsgroup] : >

: > What he said. Some (most now) power supplies are power factor corrected. : > They have filters that cause a small amount of leakage current to flow on : > the ground. Sum enough of these together and you trip a GFI (or even an RCD : > for really large gear). It appears your two machines each have power : > supplies that are close to tripping the GFI with one power supply connected, : > and will trip when both are connected. : >

: > Charles Perry P.E. : >

: >

: : I wondered about this, is there actually a leakage current or is th : current phase shifted to the extent the GFCI "thinks" there is an imbalance? : : -- : Joe Leikhim K4SAT : "The RFI-EMI-GUY"© : : "Treason doth never prosper: what's the reason? : For if it prosper, none dare call it treason." : : "Follow The Money" ;-P

Actual current.

Charles Perry P.E.

| I wondered about this, is there actually a leakage current or is th | current phase shifted to the extent the GFCI "thinks" there is an imbalance?

If the shifted phase is carried opposite on each of the 2 conductors and none of it is on the ground or leaks to an external ground, the GFCI will see it as a net zero common mode current, and do nothing.

The GFCI design is basically a single current transformer around the group of

2, 3, or 4 current carrying conductors (the number depending on the system type). The ground conductor(s) are not involved with the current transformer. What it is measuring is the sum current in common mode on all those conductors together. There cannot be such a current in common unless it forms a complete circuit somewhere else. The whole idea is that if it is forming a complete circuit somewhere else, that is a hazard (you might be part of that circuit).

Which begs the question why there would be leakage current in the design of a PF corrected UPS?

It has little or nothing to do with "PF correction."

Because switching power supplies (and, BTW, the computer itself) produce a LOT of high frequency "noise," there is usually some kind of low pass filter "network" at the power input.

The network almost always has a DC pathway between each of the power conductors and "ground"/chassis.

When one of the power conductors is at are near ground potential, the network will pass a "common mode" net current. If you have several of these supplies is parallel, you will exceed whatever threathold is set.

In my own home, I am "marginal" is this regard. I have two GFCI's "in series" (I know, I know, I will get 'a round tuit') and the upstream one trips occasionally during externally caused power surges. The potential "leakage" loads include an air conditioned, a PC on a UPS, a monitor, and a printer. The load includes an X-10 light dimmer and an outside light fixture with motion detector.

Please don't lecture me on this: it's just a case of the shoemaker's kids having no shoes! My point is that the minor "commond mode" contributions of all this crap adds up.

** Posted from **

Possibly through the EMI filters.

On Mon, 07 Jul 2008 20:29:47 +1000 ARLOWE wrote: | RFI-EMI-GUY submitted this idea : |> snipped-for-privacy@ipal.net wrote: |>> On Sun, 06 Jul 2008 12:42:01 -0400 RFI-EMI-GUY |>> wrote: |>> |>> > I wondered about this, is there actually a leakage current or is th |>> current phase shifted to the extent the GFCI "thinks" there is an |>> imbalance? |>> |>> If the shifted phase is carried opposite on each of the 2 conductors and |>> none |>> of it is on the ground or leaks to an external ground, the GFCI will see it |>> as |>> a net zero common mode current, and do nothing. |>> |>> The GFCI design is basically a single current transformer around the group |>> of |>> 2, 3, or 4 current carrying conductors (the number depending on the system |>> type). The ground conductor(s) are not involved with the current |>> transformer. |>> What it is measuring is the sum current in common mode on all those |>> conductors |>> together. There cannot be such a current in common unless it forms a |>> complete |>> circuit somewhere else. The whole idea is that if it is forming a complete |>> circuit somewhere else, that is a hazard (you might be part of that |>> circuit). |>> |>

|> Which begs the question why there would be leakage current in the design of a |> PF corrected UPS? | | Possibly through the EMI filters.

Which is even more likely with dirty (lots of noise and harmonics) power.

Maybe we need a ground fault detector design that avoids issues like this by having a high frequency bypass around the current transformer? OTOH, if 6ma is the danger level at 60 Hz, I'd think it would also be at higher frequencies as well. So such a design might not be safe.

Maybe the solution is to clean up the power _before_ it reaches the GFCI.

Not aware that PF correction generates more RF noise. The EMI filters are aiming at RF noise, not harmonics of the line frequency (that's what the PF correction is doing).

Actually, skin effect comes into play at high frequency. You can get a skin burn at 20kHz, but it's pretty difficult to actually get the current deep enough to get near vital organs. High frequency neon display tube power supplies are consequently much safer than the old line frequency step up transformers which used to be used.

On 07 Jul 2008 16:43:20 GMT Andrew Gabriel wrote: | In article , | snipped-for-privacy@ipal.net writes: |> On Mon, 07 Jul 2008 20:29:47 +1000 ARLOWE wrote: |>| RFI-EMI-GUY submitted this idea : |>|> PF corrected UPS? |>| |>| Possibly through the EMI filters. |> |> Which is even more likely with dirty (lots of noise and harmonics) power. | | Not aware that PF correction generates more RF noise. | The EMI filters are aiming at RF noise, not harmonics of | the line frequency (that's what the PF correction is doing).

Not generate noise, but rather, divert it away, and some of that to ground. This would be noise in the voltage of the supply possibly due to current harmonics of other loads elsewhere.

|> Maybe we need a ground fault detector design that avoids issues like this by |> having a high frequency bypass around the current transformer? OTOH, if 6ma |> is the danger level at 60 Hz, I'd think it would also be at higher frequencies |> as well. So such a design might not be safe. | | Actually, skin effect comes into play at high frequency. | You can get a skin burn at 20kHz, but it's pretty difficult | to actually get the current deep enough to get near vital | organs. High frequency neon display tube power supplies are | consequently much safer than the old line frequency step up | transformers which used to be used.

I've been hit with 146 MHz before. That is definitely a burn feeling.

If you just 'rewired' some things, there's a good chance you've crossed up neutrals somewhere. For example if there are two circuits feeding a string of outlets (sometimes called a mulit-wire branch circuit??), and you somehow tied the neutral wires from two different circuits together, you won't have a *short circuit*, but the small signal that is injected into the neutral by the GFCI will find a different return path and that will be enough to trip the unit.

FYI, the modern GFCI will inject a small common-mode signal onto both the hot and neutral that pass through it. If the neutral conductor is connected somehow to ground downstream, this signal will return to the service panel via the ground wire (formally known as the 'Equipment Grounding Conductor') and loop back to the supply side of the GFCI and create a trip signal. The same thing can happen if the neutral wire downstream is connected to another neutral wire on another circuit.

daestrom

On Tue, 8 Jul 2008 21:00:54 -0400 daestrom wrote: | Will wrote: |> Our computer server room outlets for our power distribution units |> (PDUs) are protected by ground fault protection. As I understand |> it, a GFP will trigger when it senses current on the ground, and when |> it does then the GFP shuts off the circuit. We tried today to |> rewire some things in our server room, and the GFP kept triggering, |> and I want help understanding possible causes. |>

|> Each server has two power supplies. In an effort to minimize |> cabling, we purchased from Dell some cables that plug into the PDU |> using Nema 5-15P, and split to two different IEC C14 plugs that go to |> the server power supplies. It appears that use of this power cord to |> two power supplies is what is causing the ground faults. We can |> cure the GFP fault simply by leaving just one of the two plugs in |> each server. Does someone have an explanation as to why the GFP |> would fault when both power supplies are served from the same plug? |>

| | If you just 'rewired' some things, there's a good chance you've crossed up | neutrals somewhere. For example if there are two circuits feeding a string | of outlets (sometimes called a mulit-wire branch circuit??), and you somehow | tied the neutral wires from two different circuits together, you won't have | a *short circuit*, but the small signal that is injected into the neutral by | the GFCI will find a different return path and that will be enough to trip | the unit.

If he crossed neutrals to each other, you don't need a signal injecting GFCI to be tripped by that. You just need enough of a load for the current to distribute 6ma through the wrong neutral.

| FYI, the modern GFCI will inject a small common-mode signal onto both the | hot and neutral that pass through it. If the neutral conductor is connected | somehow to ground downstream, this signal will return to the service panel | via the ground wire (formally known as the 'Equipment Grounding Conductor') | and loop back to the supply side of the GFCI and create a trip signal. The | same thing can happen if the neutral wire downstream is connected to another | neutral wire on another circuit.

I thought this was for detecting reversed hot/neutral or neutral/ground at the source. If neutral is crossed to ground downstream of the GFCI, it will be detected as a current imbalance in the hot and neutral pair since some of the current that should be on the neutral is going back over ground.

. How it detect either. .

It detects a cross with no load.

On Wed, 09 Jul 2008 13:20:31 -0500 bud-- wrote: | snipped-for-privacy@ipal.net wrote: |> On Tue, 8 Jul 2008 21:00:54 -0400 daestrom wrote: |> | FYI, the modern GFCI will inject a small common-mode signal onto both the |> | hot and neutral that pass through it. If the neutral conductor is connected |> | somehow to ground downstream, this signal will return to the service panel |> | via the ground wire (formally known as the 'Equipment Grounding Conductor') |> | and loop back to the supply side of the GFCI and create a trip signal. The |> | same thing can happen if the neutral wire downstream is connected to another |> | neutral wire on another circuit. |> |> I thought this was for detecting reversed hot/neutral or neutral/ground at |> the source. | . | How it detect either. | . |> If neutral is crossed to ground downstream of the GFCI, it will |> be detected as a current imbalance in the hot and neutral pair since some of |> the current that should be on the neutral is going back over ground. | . | It detects a cross with no load.

Ah, OK. That would be a plus.